Many surgeons favour total gastrectomy (TG) 'de principe' in the treatment of gastric cancer, but final demonstration of its advantage over subtotal gastrectomy (SG) is still lacking. We analysed survival after curative TG or SG within groups of patients stratified according to the main prognostic variables as found in multivariate analysis, i.e., nodal status, degree of invasion of the gastric wall, patient age and sex. Our series consisted of 361 patients treated by curative SG and 41 by curative TG, admitted to the Istituto Nazionale Tumori of Milan between 1965 and 1979. In patients with lymph node involvement survival appeared to be significantly better (P = 0.0005) after SG. However, stratifying for age it was found that the benefit was limited to patients over 60 years old. No significant difference in survival was found in the group without nodal involvement (N -) and invasion of the wall to the serosa or beyond. No statistical comparison was possible in N - groups with invasion confined to mucosa, submucosa or muscularis propria because of the small number of such patients who underwent TG. We conclude that SG still represents the standard reference operation for gastric carcinoma provided that a safe proximal margin of resection is guaranteed.